Availability estimation of wind power forecasting and ... · wind power due to the minimum output...
Transcript of Availability estimation of wind power forecasting and ... · wind power due to the minimum output...
Availability estimation of wind power forecastingand optimization of day-ahead unit commitment
Yun TENG1 , Qian HUI1,2, Yan LI3, Ouyang LENG4,
Zhe CHEN1,5
Abstract Due to the uncertainty of the accuracy of wind
power forecasting, wind turbines cannot be accurately
equated with dispatchable units in the preparation of a day-
ahead dispatching plan for power grid. A robust opti-
mization model for the uncertainty of wind power fore-
casting with a given confidence level is established. Based
on the forecasting value of wind power and the divergence
function of forecasting error, a robust evaluation method
for the availability of wind power forecasting during given
load peaks is established. A simulation example is estab-
lished based on a power system in Northeast China and an
IEEE 39-node model. The availability estimation parame-
ters are used to calculate the equivalent value of wind
power of the conventional unit to participate in the day-
ahead dispatching plan. The simulation results show that
the model can effectively handle the challenge of uncer-
tainty of wind power forecasting, and enhance the con-
sumption of wind power for the power system.
Keywords Consumption of wind power, Wind power
availability, Uncertainty of wind power forecasting,
Coordination dispatching, Robust estimation
1 Introduction
With the ongoing global petrochemical energy deple-
tion, environmental degradation is an increasingly serious
situation. The global Energy Internet, with renewable
energy, has emerged as the main energy supply form and
electric energy as the main energy carrier, which is an
inevitable trend for the development of human energy
system and energy industry [1–3]. In addition, such power
generations with renewable energy such as wind power and
photovoltaic power, have transformed the pattern of mod-
ern power system and the form of power generation [4]. In
the current pattern, one of the serious challenges is that the
operation of power grid requires stability, while the inte-
gration scale of fluctuating unpredicted renewable energies
has been increasing [5, 6].
Wind power generation is now a key component in
renewable energy generation. In 2016, the installed
capacity in China increased to 169 billion kW while the
CrossCheck date: 28 August 2019
Received: 21 November 2018 / Accepted: 28 August 2019 / Published
online: 16 November 2019
� The Author(s) 2019
& Yun TENG
Qian HUI
Yan LI
Ouyang LENG
Zhe CHEN
1 Shenyang University of Technology, Shenyang 110870,
China
2 State Grid Liaoning Electric Power Research Institute
Customer Service Center, Shenyang 110004, China
3 State Grid East Inner Mongolia Electric Power Co, Ltd.,
Huhhot 010020, China
4 Economic and Technological Research Institute of State Grid
East Inner Mongolia Electric Power Co. Ltd, Huhhot 010020,
China
5 Aalborg University, Aalborg, Denmark
123
J. Mod. Power Syst. Clean Energy (2019) 7(6):1675–1683
https://doi.org/10.1007/s40565-019-00571-5
abandoned wind power increased to 49.7 billion kW, a
50% increase over the previous year, with 17% rate of
curtailed power [7–9]. The main reason is that the output of
wind power is obviously random and uncertain due to the
climate and environment change. The increasing scale of
the existing system, including wind power, will have to
guarantee a stronger power balance in power system and
make its operation less hazardous. Therefore, the current
methods are being challenged [10–13].
Wind power forecasting and its ongoing improvements
are eye-catching in wind power industries and the man-
agement of power system. So far, the studies of wind
power forecasting and its improvement are based on the
assumption that completely accurate information of geo-
graphic environment and climate is given. Then, the
investigation of theory or method about the relationship
between this information and wind power capacity is
determined. And on this basis, the probability distribution
of errors is forecasted [14, 15]. At present, the general
method within and outside China is based on the criteria of
climate, namely, the improvement of forecasting accuracy
of wind power output. The model of the numerical weather
prediction (NWP) as well as the information about the
developer, is listed in [16]. On such a basis, the Kalman
filtering algorithm is proposed in [17] to filter the output of
the SKIRON NWP model, which significantly improves
the forecasting accuracy of wind speed and wind power.
A scenario tree tool is developed in [17], which allows
the statistics of wind power forecasting error to be altered
and facilitates the study of how these statistics impact unit
commitment and system operation. A general methodology
for deriving optimal bidding strategies based on proba-
bilistic forecasting of wind generation is proposed in [18],
and the sensitivity of wind power costs has been analyzed.
The opportunities available are explored for wind power
producers (WPPs) if they can purchase or schedule some
reserves to offset part of their deviations rather than being
fully penalized in the real-time market, and the revenue for
WPPs with such mechanisms is modeled in [19].
However, as the dispatching systems of power grid are
facing more and more uncertainties related to wind power
forecasting errors, the scheduling system of power grid has
insufficient information to handle these uncertainties under
the fast increasing scale of wind power integration/inter-
connection among various energies. Therefore, the current
method demanding accurate information cannot fulfill the
need in the real work of the scheduling system [20–23]. It
is impossible to take the forecasting value of wind power as
the effective power output in day-ahead dispatching plan
based on the reliability of power supply during the peak
load period. This results in a large number of curtailed
wind power due to the minimum output limit of conven-
tional units during a low-load period.
Therefore, wind power availability can be accurately
evaluated or calculated by using a robust estimation. The
minimum possible wind power output can be inferred
based on the forecasting results of wind power, so that
wind power can be used as a reliable source to replace a
certain proportion of conventional energy units and par-
ticipate in the day-ahead dispatching plan of power system.
To improve the efficiency of unit commitment, it is nec-
essary to reduce the number of unit start-ups and shut-
downs, and to have a decisive role in significantly reducing
the curtailed wind.
A robust estimation model for the maximum probability
rate of wind power prediction and its probability distribu-
tion is proposed, which is different from the traditional
theory of improvement of wind power forecasting accu-
racy. With the current theory of prediction of wind power,
based on the results of wind power forecasting, the avail-
ability of wind power during the peak load period can be
obtained. According to their availability, the unit com-
mitment can be better arranged to tackle the conflict
between the fluctuation of wind power and the reserve
capacity, so the challenge of dispatching faced by the
power grid with large-scale wind power can be handled.
2 Availability index of wind power forecasting
2.1 Definition of accuracy of wind power forecasting
Assume that the time series of the actual output of wind
power is xi, and the predicted counterpart is x0i. The abso-
lute value of the error of the wind power forecasting at a
future time can be expressed as:
ei ¼ x0i � xi��
�� ð1Þ
This paper defines the accuracy of wind power
forecasting at a future time as:
RA ¼ x0i � xi
x0ii ¼ ijxi\x0i
� �
ð2Þ
where RA is the accuracy of wind power forecasting.
2.2 Definition of availability of wind power
forecasting
According to the conception of the accuracy of wind
power forecasting, the availability index for wind power
forecasting is defined. Different from the evaluation of
conventional wind power forecasting error or the differ-
ence between the estimated value and the actual output, the
availability index is able to reflect the minimum possible
value of wind power during the peak load period, or the
maximum possible value of wind power forecasting error,
1676 Yun TENG et al.
123
and the possibility of the lowest wind power output based
on the forecasting results.
The definition of the availability index of wind power
forecasting is:
Q ¼ maxPWA þ PWS
PWF
� �
PWI ¼ Q � PWF
8
<
:ð3Þ
where Q is the day-ahead availability of wind power
forecasting; PWA is the actual wind power during the peak
load period; PWS is the actual curtailed wind power; PWF is
the estimated value of wind power during the day-ahead
peak load period; and PWI is the capacity of the output of
virtual wind power unit participating in the day-ahead
scheduling plan during predicted peak load period.
In the optimum unit commitment of the day-ahead dis-
patching, according to the availability Q, the value of the
wind power forecasting is equivalent to the virtual unit to
participate in the unit commitment optimization, which can
effectively reduce the number of running machines.
3 Robust estimation model of availability of windpower forecasting
3.1 Robust estimation of model during various peak
load periods
During a day-ahead dispatching period, the power grid
faces a scenario d with uncertain wind power forecasting,
so the arrangement of the day-ahead scheduling plan
requires the availability Q and its probability c given by
wind power enterprises. In the actual scenario, the output
given by the enterprise is r. After a dispatching period, if
the actual output of wind power is less than the corre-
sponding scale of availability, the power of other power
sources will be Ps. If the actual output of wind power is
more than the corresponding scale of credibility, the sur-
plus will be the amount of curtailed wind power v.
Therefore, the issues of unit commitment of day-ahead
dispatching will be equivalent to those of ensuring the
optimum credibility Q* of wind power forecasting at the
beginning of the day-ahead dispatching period. And then
the product between Q* and the results of wind power
forecasting is regarded as the maximum available power of
a controllable source, which is the same as the virtual unit
to participate in the day-ahead unit commitment plan.
With the load data, the value of wind power forecasting,
the actual output of wind power without the availability of
wind power forecasting and its statistical distribution, will
be the forecasted scenario d which contains m number of
scenarios, according to the duration of peak load period:
d ¼ d1; d2; . . .; dmf g ð4Þ
The value of each scenario di (i = 1, 2, …, m)
corresponds to one duration of the load peak. In scenario
di, the probabilities of wind power forecasting error ei(i = 1, 2, …, m) are assumed to be:
Prðd ¼ diÞ ¼ pi ð5Þ
where pi � 0 andPm
i¼1
pi ¼ 1.
In order to make the wind power forecasting credibility
Q accurately reflect the actual output of the wind power,
the fitness function is defined as:
f Q; dið Þ ¼ rmin di;Qð Þ � Psmax di � Qð Þþmax Q� dið Þ � cQ
ð6Þ
The optimal value of availability of wind power
forecasting is:
Q� ¼ argmin E f Q; dið Þ½ � ¼Xm
i¼1
pif Q; dið Þ( )
ð7Þ
where E½�� is the expectation operator.
If p = (p1, p2, …, pm)T is the undetermined probability
vector of the error of wind power forecasting sometime in
the future, then under the uncertain distribution of proba-
bility of wind power forecasting error, the robust estima-
tion of availability of wind power forecasting can be
expressed as:
maxQ;aQ
minp
aQ � 1
bE max aQ � f Q; dið Þ½ �� �
� �
¼ maxQ;aQ
minp
aQ � 1
b
Xm
i¼1
pimax aQ � f Q; dið Þ½ �( ) ð8Þ
where b is the fitness risk tolerance, which indicates that
the wind power is less than the risk probability of the
credible wind power output; and aQ is the threshold of the
fitness function n(Q;d).By introducing the assistant vector u = (u1, u2, …, um)
T,
(5) is transformed into the robust optimization:
max #Q;aQ;#;u
s:t: minp
aQ � 1
bpTu�#
ui � aQ � f Q; dið Þ i ¼ 1; 2; . . .;mui � 0 i ¼ 1; 2; . . .;m
8
>>>>><
>>>>>:
ð9Þ
where (Q, aQ, #, u)[R 9 R 9 R 9 Rm; and # is the aux-
iliary parameter.
Availability estimation of wind power forecasting and optimization of day-ahead unit commitment 1677
123
3.2 Solution of robust estimation model
Since (9) includes the undetermined probability vector
p, the direct solution is hardly known. In this paper, the
above optimum is transformed into the estimation of the
confidence region of wind power forecasting availability to
fulfill the given confidence level. Therefore, the optimum
issue of (9) is turned into the issue of robust planning.
Assume that the probability vector of wind power fore-
casting error during every peak load period is p = (p1, p2,…,
pm)T, wherepi (i = 1, 2,…,m) C 0, and the probability vector
of availability of wind power forecasting in the scenarios is
q = (q1, q2, …, qm)T, where qi (i = 1, 2, …, m) C 0. The
divergence function between the forecasting availability and
probability of forecasting error is / [24], then:
I/ p; qð Þ ¼Xm
i¼1
qi/pi
qi
� �
ð10Þ
where I/ p; qð Þ is the divergence function of vectors p and
q.
Suppose that the forecasted result samples are N, then
under the confidence level of 1 - K where K is the sig-
nificance level, the confidence region of the probability p
of the wind power prediction availability under each sce-
nario is:
lN ¼ p 2 Rm pi � 0; eTp ¼ 1; I/ p; p̂Nð Þ�� � q
� �
ð11Þ
where q is auxiliary parameters; eT = (e1, e2, …, em) is the
error vector under every scenario; p̂N ¼p1;N ; p2;N . . .; pm;N
is the maximum likelihood estimation
of probability p based on N samples of a small scale:
q ¼ /00ð1Þ2N
ffiffiffiffiffi
d/p
v2m�1;1�K þ c/�
ð12Þ
where v2m�1;1�K is the value of the chi-square function of
m - 1 under the confidence interval of 1 - K; d/ and c/are correction parameters defined in (13) and (14),
respectively; and /00, /ð3Þ and /ð4Þ are the second, third and
fourth derivatives of /ðtÞ ¼ 1tðt � 1Þ2, respectively.
d/ ¼ 1þ 1
2 m� 1ð ÞN 2� 2m� m2 þ s
!(
þ 2/ð3Þ 1ð Þ/00 1ð Þ 4� 6m� m2 þ 3s
þ 1
3
/ð3Þ 1ð Þ/00 1ð Þ
!2
4� 6m� m2 þ 5s
þ 2/ð4Þ 1ð Þ/00 1ð Þ 1� 2mþ sð Þ
)
ð13Þ
c/ ¼ m� 1ð Þ 1�ffiffiffiffiffi
d/p�
þ 1
N
/ð3Þ 1ð Þ3/00 1ð Þ 2� 3mþ sð Þ þ /ð4Þ 1ð Þ
4/00 1ð Þ ð1� 2mþ sÞ !
ð14Þ
where s is the auxiliary parameter and s ¼Pm
i¼1
1pi;N
.
Then, under the uncertain probability set, the minimum
constraint problem of equation (9) is turned into:
aQ � g� kq� kPm
i¼1
pi;N/� li=b� g
k
� �
�#
k� 0
8
<
:ð15Þ
where /�ð�Þ is the conjugate function of /ð�Þ; g and k are
auxiliary parameters. The issue about parameter
optimization in equation (6) can be turned into the one
concerning robust planning:
maxQ;#;k;g;u
#
s:t: aQ � g� kq� kPm
i¼1 pi;N/� li=b� g
k
� �
�#
ui � aQ � f Q; dið Þ i ¼ 1; 2; . . .;mui � 0 i ¼ 1; 2; . . .;mk� 0
8
>>>>>>><
>>>>>>>:
ð16Þ
where (Q, #, k, l, u) [ R 9 R 9R 9 R 9 Rm.
4 Estimation of algorithm of wind powerforecasting availability and application analysis
4.1 Robust estimation algorithm of wind forecasting
availability
Based on load data, wind power prediction data, actual
wind power data and wind curtailment data in Northeast
China from 2014 to 2016, a u-divergence robust estimation
model concerning the availability of wind power fore-
casting has been built. Based on wind power forecasting
data used in day-ahead dispatching, the availability of wind
power forecasting is estimated.
The typical daily load curves of the power grid of the
region in winter and summer are shown in Fig. 1.
The time of the daily maximum load of the system and
the distribution of errors of wind power forecasting are
shown in Figs. 2 and 3, respectively.
According to the analysis, the number of scenarios of
the peak load duration corresponding to wind power fore-
casting error is m = 10. From the historical data analysis,
the actual probability of wind power forecasting error is p0in each scenario of the regional power grid. Parameters
related to the robust optimization model of availability of
1678 Yun TENG et al.
123
wind power forecasting are: r = 0.12, Ps = 0.08, v = 0.1,
c = 0.92. Set ei = 10%, and the distance function is v2
according to divergence. The robust estimation algorithm
of the availability of wind power forecasting in the next
period of day-ahead dispatching of the local power grid is
shown in Fig. 4, and is explained as follows:
1) According to durations of the load peaks of the power
grid, the scenario mode is d = (1, 2, 5, 10, 15, 18, 20,
25, 30, 60), in terms of minutes. Based on the
historical forecasting data, the probability vector with
the wind power forecasting error ei = 10% in each
scenario is p0 = (0.80, 0.82, 0.85, 0.81, 0.93, 0.85,
0.91, 0.93, 0.91, 0.93)T.
2) Scenario samples with a capacity of N = 36, 110, 156,
365, 548, 1095 are obtained, corresponding to the load
of one month, ten days, one week, three days, two days
and one day, respectively. 100 samples are used to
calculate the mean value.
3) As the samples are of different scales, based on the
maximum likelihood estimation method, the sample-
based probability in each peak load scene is pi,N (i = 1,
2, …, m).
4) When the distance divergence function and confidence
level (1 - K) are 95% and 99%, respectively, and the
fitness and risk tolerance b are 0.03 and 0.09,
respectively, the solution of (15) is calculated to
1 4 7 10 12Month
00:00
06:00
12:00
18:00
24:00
Tim
e
29 October1 May
2 3 5 6 8 9 11
Fig. 2 Periods of daily maximum load of system
0 50 100 150 200 250 300 350
Time (day)
0.5
1.0
1.5
2.0
2.5
3.0
Win
d po
wer
fore
cast
ing
accu
racy
(%)
Maximum error is 0.63%
400
Fig. 3 Statistical distribution of accuracy of daily wind power
forecasting
Obtain the data of load forecasting and wind power forecasting
Obtain specific scenarios of load peaks
Set the value of risk tolerance
Calculate the reliability of wind power output of the specific load peak duration
Calculate the reliable output of wind power with the reliability
Optimizeday-ahead unit commitment
Is the output of total unit less than the risk of peak (with the given
uncertainty of load)?
Output unit commitmentY
N
Start
End
Fig. 4 Robust estimation flowchart of wind forecasting availability
00:00 06:00 12:00 18:00 24:00Time
12000
13000
14000
15000
16000
17000Po
wer
(MW
) Typical daily load in summer
Typical daily load in winter
Fig. 1 Curves of typical daily load of system
Availability estimation of wind power forecasting and optimization of day-ahead unit commitment 1679
123
obtain the robust availability of the wind power
forecasting.
Based on the forecasted value of wind power obtained
from day-ahead dispatching, by the robust estimation
model of wind power forecasting availability proposed in
this study, the distribution of wind power forecasting
availability with load peak duration of 15 min on the next
day is calculated as shown in Fig. 5.
4.2 Analysis of calculation of day-ahead dispatching
based on wind power forecasting
Based on the MATLAB simulation platform, this paper
adopts the wind power forecasting credibility robust eval-
uation method when the load peak scenario appears. Tak-
ing for example a 39-node power system with 8000 MW
total capacities shown in Fig. 6, all the wind farms in the
power grid are equivalent to one virtual power generation
unit, and all other power sources are equivalent to a con-
ventional thermal unit. The regulating capacity of the vir-
tual wind power unit is set to 100%, and the regulating
capacity of the conventional thermal power unit is set to
50%. In Fig. 6, ‘‘W’’ represents wind turbine.
The estimated curves of the day-ahead load and wind
power forecasting are shown in Fig. 7. The grid-connected
capacity of the wind farm takes 10% and 20% of the total
capacity of the system, and the calculation time is 12.8 s
and 14.2 s, respectively, which is used to analyze the
effectiveness of the model under different wind power
conditions.
The model during the load peak of the power grid is
d ¼ f1; 2; 5; 10; 15; 18; 20; 25; 30; 60g. According to the
data of wind power forecasting, if the error of day-head wind power forecasting is xi=10%, the probability of the
day-ahead error under each period is p0 ¼ 0:80; 0:82;ð0:85; 0:81; 0:93; 0:85; 0:91; 0:93; 0:91; 0:93ÞT:
According to the above results, by the robust estimation
model of wind power forecasting availability, the day-head
availability based on the scenarios of b = 0.03 and
b = 0.09 is calculated, respectively. According to the cal-
culated index of wind power availability, the maximum
output power of the available virtual wind power units
during the peak load of the power grid is calculated. The
results are shown in Table 1.
According to the adjustable peak output value of the
virtual wind power units in Table 1, the virtual units and
the conventional thermal power units in the system are
taken as the units with the same attribute for day-ahead
dispatching. Unit commitment is shown in Table 2. Fig-
ure 8 shows the relationship between risk tolerance and
availability.
The capacity of the connection of the power grids of the
wind power field is 20% of the total capacity of the system,
0 0.5 1.0 1.5 2.0 2.5 3.0
Wind power forecasting availability (%)
5
10
15
20
25
30
35
Freq
uenc
y
Credibleaccuracy
40
Fig. 5 Availability of wind power forecasting
00:00 06:00 12:00 18:00 24:00Time
02000400060008000
10000120001400016000 Day load forecasting
Case 1: wind power forecasting
Case 2: wind power output
Pow
er (M
W)
Fig. 7 Day-ahead load and wind power forecasting curves
G1 G8
G10 G9
G3G2 G5 G4
G7
G6
30 3725
21
39
3
45
9
8
711
1213
31 3210
34 33
2019
14
1516 21
17
2726 28 29
38
3624
23
2235
6
18
W
Fig. 6 Equivalent system diagram of power grid in one region
1680 Yun TENG et al.
123
as we can see from Table 2. The product between the value
of wind power forecasting and the index of credibility is
equivalent to one unit, so the current number of units is
reduced by 1 and 2 with the two risk tolerances, respec-
tively. The number of running thermal power units is
decreased.
The actual load of the power grid and the wind power
output are assumed to be as shown in Fig. 9.
The simulation is proceeded with the two day-ahead unit
commitment strategies of b = 0.03 and b = 0.09,
respectively. The capacity of the connection of wind power
accounts for 20% of the total capacity of the system. The
unit commitment is equivalent to the availability of wind
power forecasting, which is unconsidered. The counterpart
is proceeded when the unit commitment policy of day-
ahead dispatching, as shown in Table 2, is considered.
Due to the uncertainty of wind power which is not
involved in the units, the conventional day-ahead dis-
patching plan is hardly known to optimize the number of
conventional thermal units and ensure the stability of the
source power during the peak load period. Therefore,
during the valley load period, the minimum output of the
thermal unit occupies the room of wind power consump-
tion, and curtailed wind is formed. As shown in Figs. 10
and 11, the adaption of the index of wind power forecasting
makes the wind power consumption much better than that
shown in Fig. 12, and hence the ratio of curtailed wind
power of the grid power is reduced. The larger the risk
tolerance is, the better wind power consumption will be. As
a result, the reasonable index of risk tolerance can make the
relationship controllable between wind power consumption
and safe operation of power grid.
Table 1 Optimal confidence and condition risk value with different
risk tolerances
Condition Optimal
availability
Adjustable peak output of wind
power virtual unit (MW)
Wind power
ratio is 10%,
b is 0.03
0.64 150
Wind power
ratio is 10%,
b is 0.09
0.68 163
Wind power ratio is
20%, b is 0.03
0.61 286
Wind power
ratio is 20%,
b is 0.09
0.65 312
Table 2 Day-ahead unit commitment of dispatching
Risk tolerance b Day-ahead unit commitment
0.03 W, G1, G2, G4, G6, G7, G8, G9, G11, G12, G15,
0.09 W, G1, G2, G3, G5, G7, G8, G10, G12, G14
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.160.54
0.56
0.58
0.60
0.62
0.64
0.66
0.68
0.70
0.72
Q
β
Case 1Case 2
Fig. 8 Relationship between risk tolerance and availability
Actual load
Case 1: wind power outputCase 2: wind power output
02000400060008000
10000120001400016000
Pow
er (M
W)
00:00 06:00 12:00 18:00 24:00Time
Fig. 9 Actual load and wind power output
06:00 12:00 18:00 24:00Time
Pow
er (M
W) t=18:00
2000
1500
1000
500
0
50% capacity
00:00
Consumption of wind power
0.650.03, *== Qβ
Virtual generator capacityForecasted wind power; Curtailed wind power
Fig. 10 Wind power consumption with day-ahead dispatching policy
based on their availability when b = 0.03
Availability estimation of wind power forecasting and optimization of day-ahead unit commitment 1681
123
5 Conclusion
In this paper, the availability index of wind power
forecasting and the model of virtual wind power unit
commitment with the minimum value of total wind power
output during a selected period are proposed, based on the
characteristics of historical data of the running power grid
and the index of wind power forecasting evaluation.
Wind power forecasting credibility and its robust esti-
mation model of probability distribution are proposed. The
solution algorithm based on the robust plan is
established.
The equivalent stimulating mode of 39-node power grid
in 10 units is established. Based on the estimated model of
the availability of wind power forecasting, the unit com-
mitment is optimized. The simulation results show that the
robust estimation of availability of wind power forecasting
and the optimization of the unit commitment can effec-
tively improve the wind power consumption of the power
grid.
Acknowledgements This work was supported by the National Key
Research and Development Program of China (No.
2017YFB0902100).
Open Access This article is distributed under the terms of the
Creative Commons Attribution 4.0 International License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a
link to the Creative Commons license, and indicate if changes were
made.
References
[1] Bruninx K, Delarue E (2014) A statistical description of the
error on wind power forecasts for probabilistic reserve sizing.
IEEE Trans Sustain Energy 5(3):995–1002
[2] Shil SK, Sadaoui S (2018) Meeting peak electricity demand
through combinatorial reverse auctioning of renewable energy.
J Mod Power Syst Clean Energy 6(1):73–84
[3] Liu F, Bie Z, Liu S (2017) Day-ahead optimal dispatch for wind
integrated power system considering zonal reserve require-
ments. Appl Energy 188:399–408
[4] Teng Y, Sun P, Hui Q et al (2019) Multi-energy microgrid
autonomous optimized operation control with electro-thermal
hybrid storage. CSEE J Power Energy Syst. https://doi.org/10.
17775/cseejpes.2019.00220
[5] Zhang N, Kang C, Xia Q (2014) Modeling conditional forecast
error for wind power in generation scheduling. IEEE Trans
Power Syst 29(3):1316–1324
[6] Zhao L, Wei G, Zhi W et al (2018) A robust optimization
method for energy management of CCHP microgrid. J Mod
Power Syst Clean Energy 6(1):132–144
[7] Zarate-Minano R, Anghel M, Milano F (2013) Continuous wind
speed models based on stochastic differential equations. Appl
Energy 104:42–49
[8] Rajakovic NL, Shiljkut VM (2018) Long-term forecasting of
annual peak load considering effects of demand-side programs.
J Mod Power Syst Clean Energy 6(1):145–157
[9] Chang WY (2014) A literature review of wind forecasting
methods. J Power Energy Eng 2(04):161–168
[10] Liu L, Ji T, Li M et al (2018) Short-term local prediction of
wind speed and wind power based on singular spectrum analysis
and locality-sensitive hashing. J Mod Power Syst Clean Energy
6(2):317–329
[11] Li H, Wang J, Lu H (2018) Research and application of a
combined model based on variable weight for short term wind
speed forecasting. Renew Energy 116:669–684
[12] Chen H, Li F, Wang Y (2018) Wind power forecasting based on
outlier smooth transition autoregressive GARCH model. J Mod
Power Syst Clean Energy 6(3):532–539
[13] Sedighi M, Moradzadeh M, Kukrer O et al (2018) Simultaneous
optimization of electrical interconnection configuration and
cable sizing in offshore wind farms. J Modern Power Syst Clean
Energy 6(4):749–762
[14] Wang J, Du P, Niu T (2017) A novel hybrid system based on a
new proposed algorithm—multi-objective whale optimization
algorithm for wind speed forecasting. Appl Energy
208:344–360
[15] Zhang Z, Mei D, Jiang H et al (2018) Mode for reducing wind
curtailment based on battery transportation. J Mod Power Syst
Clean Energy 6(6):1158–1171
[16] Wang J, Niu T, Lu H et al (2018) An analysis-forecast system
for uncertainty modeling of wind speed: a case study of large-
scale wind farms. Appl Energy 211:492–512
06:00 12:00 18:00 24:00Time
Pow
er (M
W) t=18:00
2000
1500
1000
500
0
50% capacity
00:00
0.610.09, *== Qβ
Virtual generator capacityForecasted wind power; Curtailed wind power
Consumption of wind power
Fig. 11 Wind power consumption with day-ahead scheduling policy
based on their availability when b = 0.09
06:00 12:00 18:00 24:00Time
Pow
er (M
W)
2000
1500
1000
500
000:00
Forecasted wind power; Curtailedwind power
Fig. 12 Curtailed wind power with day-ahead scheduling policy of
virtual unit ignoring their availability
1682 Yun TENG et al.
123
[17] Yuan K, Zhang K, Zheng Y et al (2018) Irregular distribution of
wind power prediction. J Mod Power Syst Clean Energy
6(6):1172–1180
[18] Chen F, Huang G, Fan Y et al (2017) A copula-based fuzzy
chance-constrained programming model and its application to
electric power generation systems planning. Appl Energy
187:291–309
[19] Jiang Y, Chen X, Yu K et al (2017) Short-term wind power
forecasting using hybrid method based on enhanced boosting
algorithm. J Mod Power Syst Clean Energy 5(1):126–133
[20] Quan H, Srinivasan D, Khambadkone AM et al (2015) A
computational framework for uncertainty integration in
stochastic unit commitment with intermittent renewable energy
sources. Appl Energy 152:71–82
[21] Basit A, Hansen A, Sørensen P et al (2017) Real-time impact of
power balancing on power system operation with large scale
integration of wind power. J Mod Power Syst Clean Energy
5(2):202–210
[22] Wang Y, Zhang N, Kang C (2018) An efficient approach to
power system uncertainty analysis with high-dimensional
dependencies. IEEE Trans Power Syst 33(3):2984–2994
[23] Xie Y, Liu C, Wu Q (2017) Optimized dispatch of wind farms
with power control capability for power system restoration.
J Mod Power Syst Clean Energy 5(6):908–916
[24] Doostizadeh M, Aminifar F, Ghasemi H et al (2016) Energy and
reserve scheduling under wind power uncertainty: an
adjustable interval approach. IEEE Trans Smart Grid
7(6):2943–2952
Yun TENG received his Ph.D. degree in electrical engineering from
Shenyang University of Technology, Shenyang, China, in 2009. He
joined Shenyang University of Technology in 2010, where he is
currently a Professor in electrical engineering. He is a standing
director of the IEEE PES DC Distribution Network Technical
Subcommittee. His research interests include multi-energy system
dispatching automation and smart grid control theory.
Qian HUI received her M.Sc. degree in electrical engineering from
Shenyang University of Technology, Shenyang, China, in 2017. She
is currently pursuing her Ph.D. degree at the faculty of Electrical
Engineering, Shenyang University of Technology. She joined the
State Grid Liaoning Electric Power Research Institute Customer
Service Center, China, in 2017. Her major research interests include
new energy power systems and multi-energy coordinated dispatching.
Yan LI received his Ph.D. degree in electrical engineering from
Shenyang University of Technology, Shenyang, China, in 2018. He
joined the State Grid East Inner Mongolia Electric Power Supply Co.
Ltd., China, in 2009, where he is currently an Electrical Engineer. His
research interests include ultra-high voltage (UHV) substations,
power transmission project construction, and power system
simulation.
Ouyang LENG received her B.S. degree from North China Electric
Power University, Beijing, China, in 2008. She joined the Economic
and Technological Research Institute of the State Grid East Inner
Mongolia Electric Power Co. Ltd., China, in 2016. Her major research
directions include new energy power system and coordinated
dispatching.
Zhe CHEN received his B.Eng. And M.Sc. degrees from Northeast
China Institute of Electric Power Engineering, Changchun, China,
and his Ph.D. degree from the University of Durham, Durham, U.K.
He is a Full Professor with the Department of Energy Technology,
Aalborg University, Aalborg, Denmark, where he is the leader of the
Wind Power System Research program in the Department of Energy
Technology. He is also the Danish Principle Investigator for Wind
Energy of the Sino-Danish Centre for Education and Research. His
research areas are power systems, power electronics and electric
machines, and his main current research interests are wind energy and
modern power systems. He has led many research projects and has
authored or coauthored more than 400 publications in his technical
field. Dr. Chen is a Fellow of the Institution of Engineering and
Technology and a Chartered Engineer in the U.K. He is an editor of
IEEE Transactions on Power Systems, and an associate editor of
IEEE Transactions on Power Electronics and Journal of Modern
Power Systems and Clean Energy.
Availability estimation of wind power forecasting and optimization of day-ahead unit commitment 1683
123